Researchers from the Beijing Institute of Technology, Peking University and Japan's National Institute for Materials Science have developed a few-layer graphene/CrOCl/few-layer graphene van der Waals (vdW) heterostructure that functions as a broadband infrared optoelectronic synaptic device, notable for its tunable spike timing-dependent plasticity and a broad spectral response range of 520–2000 nanometers. This system addresses the limitations of conventional 2D synaptic devices, which are typically restricted to visible light due to their material bandgaps.
The device is composed of two thin graphene layers separated by a chromium oxychloride (CrOCl) barrier, all encapsulated in hexagonal boron nitride for stability. When a voltage is applied, electrons tunnel through the CrOCl layer, even in the absence of light—a behavior distinct from metal-contact devices. Upon illumination (including infrared wavelengths), interfacial coupling drives charge transfer from graphene to CrOCl, modulating the tunneling barrier and triggering synaptic plasticity effects, such as spike-number and frequency-dependent plasticity.
Quantum mechanical calculations (density functional theory) support that photoexcited charge transfer shifts CrOCl’s conduction band, reducing the tunneling barrier and facilitating current increase. Experiments reveal that when illuminated, the barrier is modulated and the device exhibits persistent photoresponse memory, analogous to biological synapse function.
The optical synaptic device demonstrates efficient simulation of multiple forms of synaptic plasticity, including spike-timing-dependent plasticity (STDP), spike-number-dependent plasticity (SNDP), spike-rate-dependent plasticity (SRDP), and paired pulse facilitation (PPF), across both visible and infrared spectral ranges (520, 1064, 1400, 1550, 2000 nm). Notably, paired-pulse facilitation at 1550 nm reached a facilitation index of ~1.84, showing adaptive properties akin to biological neural networks.
CrOCl layer thickness (2.5–40 nm) does not critically impact the synaptic behavior, indicating the response is governed predominantly by graphene–CrOCl interfacial physics rather than bulk material properties.
By employing reservoir computing, the device can process time-dependent optical signals. Infrared images encoded into pulse sequences at 1550 nm were successfully classified with an accuracy exceeding 98%, demonstrating the device’s feasibility for integration with intelligent neuromorphic systems and retinomorphic computing platforms.
The heterostructure maintained robust performance over repeated cycles and long-term operation, addressing stability issues seen in earlier infrared-sensitive materials (e.g., black phosphorus). Control experiments using hexagonal boron nitride barriers did not produce synaptic plasticity, confirming the unique role of the graphene–CrOCl interface.
This research establishes vdW heterostructures containing graphene and CrOCl as a promising foundation for next-generation broadband infrared detection, with integrated synaptic and memory processing capabilities suitable for advanced imaging, real-time sensing, and autonomous systems.
